Open zhangpengshan opened 5 years ago
Current implementation only one learning rate for both wide and deep but the speed for both layer is different, we should support two learning rate.
For example, my testing in my data set wide layer is 5 learning rate but deep layer is 0.6.
Current implementation only one learning rate for both wide and deep but the speed for both layer is different, we should support two learning rate.
For example, my testing in my data set wide layer is 5 learning rate but deep layer is 0.6.